Gateskeeper

Bill Gates has now had a chance to sound off on AI technology. He is a semi-doomer. He believes that AI will produce abundance but that the transition will feature economic calamities.

I don’t see any evidence produced for this. I’m more interested in his scoffing about regulation. He notes that we don’t just trust drugmakers to release drugs, hoping only that liability will protect consumers, then he posits that AI’s sheer power as a technology requires more.

Maybe it will, but it is not clear to me that we know what the dangers will be. The HuggingFace attack was not a truly novel event but a badly designed cybersecurity test environment. Generally, our common law is good for regulating industries especially new ones, or ones where we don’t quite know what the dangers will be. Having scared yourself with the blog LessWrong and its visions of AI-driven extinction doesn’t mean the rest of the country has to effectively appoint Anthropic and OpenAI as the only qualified custodians of this technology.

Generally, we see something like what we saw today — young industries committing to a set of professional standards and norms. Something like HAZOP in the chemical industry. Only later, as we see the actual shape and cost of accidents, will we have a decent grasp of where regulation may be needed, if at all.